Review of pool boiling enhancement by surface modification

Highlights

• This review explores passive enhancement of pool boiling using surface modification.

• Examined are macrosacle, microscale, nanoscale and multiscale (hybrid-scale) techniques.

• Addressed are performance goals of inhibiting incidence hysteresis and increasing both nucleate boiling heat transfer coefficient and CHF.

• Shown is that micro and nano surface features are susceptible to blockage and changes in performance over time.

Abstract

This paper provides a comprehensive review of published articles addressing passive enhancement of pool boiling using surface modification techniques. They include macroscale, microscale, and nanoscale surfaces, as well as multiscale (hybrid-scale), and hybrid-wettability techniques. Different enhancement methods are assessed in terms of underlying fluid routing mechanisms and ability to achieve three distinct heat transfer goals: eliminating incipient boiling hysteresis, increasing nucleate boiling heat transfer coefficient, and ameliorating critical heat flux (CHF), especially for inert dielectric coolants that are both highly wetting and possess relatively poor thermophysical properties. While different enhancement scales are shown to provide different degrees of success in achieving the three goals, it is shown that both microscale and nanoscale surface features are susceptible to blockage, resulting in deterioration of the enhancement over time. This review also points to scarcity of sufficiently sized databases for a given enhancement scheme in terms of fluid type, surface material, size, and orientation, enhancement shape, pattern, and scale, and operating pressure. This renders available findings less-than-adequate tools for design of practical cooling systems.

Introduction

Performance of devices in many applications is becoming increasingly dependent on the ability to dissipate large amounts of heat while maintaining material temperatures below prescribed limits. These include supercomputers, computer data centers, hybrid vehicle power electronics, heat exchangers for hydrogen storage, advanced radar, X-ray medical equipment, aircraft, satellite and spacecraft avionics, and laser and microwave directed energy weapons, collectively categorized as relatively low-temperature applications [1], [2], [3], [4]. But high heat dissipation is also encountered in high-temperature applications, including air-fuel heat exchangers for high-Mach aircraft engines, fusion reactor blankets, particle accelerator targets, magnetohydrodynamic (MHD) electrode walls, and rocket nozzles [1], [2], [3], [4]. They also include quenching of metal alloy parts from very high temperatures to achieve optimum alloy microstructure and superior mechanical properties [5], [6]. Rapid developments in both the low-temperature and high-temperature applications have greatly exasperated heat dissipation at device, module and system levels, which is a key reason behind recent efforts to develop better cooling schemes, especially for the low-temperature applications.

An early popular scheme to tackle the increases in heat dissipation, especially for electronic and power devices, is use of air-cooled heat sink attachments. With time, there have been many efforts to enhance the performance of these heat sinks, including use of fans incorporated into the heat sink itself and improving conductivity of interface materials. But, because of poor thermal transport properties of air, these improvements fell short of the ability to tackle increases in heat dissipation from modern devices. And, by the early 1980s, aggressive miniaturization of electronic components within high performance devices rendered air-cooled heat sinks virtually obsolete, causing a drastic shift to cooling schemes utilizing single-phase dielectric liquids. With their superior thermophysical properties compared to air, these liquids were better able to capture the dissipated heat in the form of a rise in the coolant’s sensible heat.

However, by the mid-1980s, heat dissipation in many electronic and power devices began to cross the 100 W/cm² threshold, exceeding the capabilities of most single-phase liquid cooling schemes. Since then, cooling system developers have shifted focus to two-phase cooling schemes. With their ability to capitalize upon the coolant’s sensible as well as latent heat, two-phase schemes offer the ability to dissipate far greater amounts of heat while simultaneously maintaining acceptably low device temperatures. Fig. 1 compares ranges of heat transfer coefficients attainable with different fluids and cooling schemes.

To date, several two-phase cooling schemes have been recommended for high heat flux removal. And developing fundamental understanding of the phase change mechanisms and practical implementation of these schemes have been two primary goals of studies at the Purdue University Boiling and Two-Phase Flow Laboratory (PU-BTPFL) dating back to 1984. These schemes can be grouped into passive, semi-passive, and active categories, depending on the manner in which the coolant is supplied to the heat-dissipating surface.

Passive two-phase cooling schemes rely on buoyancy-driven pool boiling. Fig. 2 shows a schematic of a typical pool boiling thermosyphon, wherein heat-dissipating modules submerged in a liquid pool release the heat via nucleate boiling. This system relies entirely on buoyancy to maintain coolant motion within a closed chamber, where vapor produced by the boiling rises to the top portion of the chamber, condenses upon the surfaces of an air-cooled or liquid-cooled condenser, and the condensed liquid drips back to the liquid pool below. Proponents of pool boiling point to several advantages: passive fluid circulation, simplicity of construction, ease of fabrication and sealing, maintenance-free operation, and absence of pump-induced fluid pulsation [7].

Semi-passive two-phase cooling schemes rely in part on buoyancy, but require a relatively small pump to assist coolant circulation. The most common variety of these schemes involves cooling with the aid of a gravity-driven (falling) liquid film from an elevated plenum, and relying on buoyancy to drive the vapor upwards, where it condenses into liquid in the plenum [9], [10], [11]. The primary purpose of the small pump is to maintain steady level of liquid in the inlet plenum. While semi-passive schemes do offer certain cooling benefits, their advantages relative to thermosyphons are somewhat limited.

Active two-phase cooling schemes rely on a pump to tackle coolant circulation within the thermal management system. They include channel flow boiling [12], [13], [14], mini/micro-channel heat sinks [15], [16], jet-impingement [17], [18], and spray cooling [19], [20], [21]. They also include hybrid cooling schemes, which combine the merits of two or more active cooling schemes in pursuit of performance superior to those of the individual schemes [22], [23]. However, the need for a large pump, complex plumbing and flow control, and high cost are all key concerns when considering implementation of any active cooling scheme.

Early liquid thermal management systems used water in a variety of ‘indirect cooling’ schemes, where the heat dissipating device is attached to a thermally conducting substrate that is subjected to the water cooling, buffering the device from any water contact. Here, thermal cooling performance is greatly influenced by thermal resistance between the device and substrate. Examples of indirect water cooling include situations involving extremely high heat fluxes, such as fusion reactor blankets [24], [25] and X-ray devices [26].

Clearly, better cooling may be realized with ‘direct cooling’ (also termed immersion cooling), where the device is subjected directly to the coolant, thereby eliminating the thermal conduction resistance. However, for current carrying electronic and power devices, the coolant in a direct cooling configuration must be both inert and dielectric. This implies that, despite its superior thermophysical properties, water cannot be used in these situations because of poor dielectric properties.

Acceptable coolants must be dielectric, inert, stable, non-flammable, and non-reactive. While Freon type refrigerants do possess these properties and have been used in the past in direct cooling situations, most are presently banned from use because of environmental concerns. Alternatives include 3 M company’s ‘Fluorinert’ and ‘Novec’ fluids, which provide superior environmental properties (especially the Novec series), and are therefore highly favored for direct cooling of electronic and power devices. However, these fluids are not without shortcomings. Key among those are relatively poor thermophysical properties, high air solubility, and extremely small contact angle. Lacking the attractive thermophysical properties of water, these fluids necessitate the use of surface enhancement techniques to meet the cooling requirements of modern electronic devices. High air solubility (e.g., 48% by volume at 1 atm and 25 °C for 3 M’s FC-72) [27] can result in artificial boiling incipience at temperatures well below saturation temperature of the pure fluid [28], [29]. These coolants must therefore be thoroughly deaerated before being charged into a tightly sealed cooling system. And very small contact angles (typically below 10°) contribute to irregularities in boiling initiation and high tendency for leaks.

The surface temperature required to initiate nucleate boiling is often determined by performing a force balance for a spherical bubble forming at a surface cavity, and relating fluid pressure to fluid temperature using the Clausius-Clapeyron equation,

$$\mathrm{\Delta }{T}_i=T_i-T_{\mathit{sat}}=\frac{T_{\mathit{sat}}{\upsilon }_{\mathit{fg}}}{h_{\mathit{fg}}}\frac{2\sigma }{r},$$

where T_i, T_sat, and r are the incipient boiling surface temperature, saturation temperature, and bubble radius, respectively. Most popular incipience theories, such as those of Hsu [30], Han and Griffith [31], and Davis and Anderson [32], are based on the assumptions of (1) existence of a relatively large vapor embryo within the cavity, and (2) that bubble radius at the mouth of the cavity is equal or proportional to the cavity radius. According to Eq. (1), this implies that the superheat required to induce bubble growth and release from a particular cavity is inversely proportional to the cavity radius. However, for dielectric fluorochemicals, very small contact angle causes appreciable penetration of liquid deep into the cavity, allowing initial capture of only a small embryo whose radius is much smaller than that of the cavity. This implies that the radius that needs to be employed in Eq. (1) should be that of the small embryo, which results in an incipience superheat much larger than predicted using the cavity radius [29]. An important factor influencing formation and size of initial embryo is the relative between the magnitude of effective cone angle, θ, of the cavity and contact angle, α. Fig. 3 shows three possible scenarios related to embryo capture as bulk liquid attempts to replenish a cavity. Combination of a relatively wide-angle cavity and smaller contact angle (θ > α), Fig. 3(a), causes the liquid to ‘flood’ the cavity, preventing formation of a vapor embryo. Fig. 3(b) shows capture of a relatively large embryo for θ < α in relatively large contact angle fluids such as water. On the other hand, Fig. 3(c) depicts the capture of a very small embryo for θ < α in very small contact angle fluids such as dielectric fluorochemicals. Growth of a very small embryo as depicted in Fig. 3(c) requires much higher superheat than predicted according to Eq. (1) using the cavity radius. For example, Reeber and Frieser [33] reported that nucleation in FC-72 did not begin even with a surface temperature 46 °C higher than saturation temperature. You et al. [34] examined the boiling incipience for FC-72 and R-113 on sputtered surfaces of various materials with cavities on the order of 1 μm. Based on measured incipience superheats of 18–51 °C for FC-72 and 43–75 °C for R-113, and using the bubble growth criterion in Eq. (1), they estimated embryo radii of 0.024–0.15 μm for FC-72, and 0.022–0.077 μm for R-113, much smaller than cavity radius. Hence, estimating boiling incipience temperature for highly wetting fluids requires determination of distribution of embryo sizes rather than those of surface features [8].

As discussed in the previous section, boiling incipience in low contact angle fluorochemicals is delayed to high surface superheats. But once incipience takes effect, this excess superheat is quickly dissipated by vigorous boiling. Additionally, bubble growth from one cavity can extend into neighboring ones, activating those cavities as well, and causing the boiling to spread rapidly over a large portion of the heated surface. These effects cause rapid transition from relative poor heat transfer dominated by natural convection in liquid just prior to boiling incipience, to highly effective nucleate boiling. The outcome of this transition is a sudden drastic decrease in surface temperature as shown in Fig. 4. The large superheat required to initiate the boiling process when increasing the surface heat flux (heating mode) is commonly referred to as ‘incipience excursion’ or ‘incipience overshoot,’ and the ensuing temperature drop immediately following incipience as ‘incipience temperature drop’. Notice how decreasing the wall heat flux (cooling mode) results in nucleate boiling characteristics virtually identical to those of the heating mode at both high fluxes corresponding to fully developed boiling, and very low heat fluxes associated with liquid natural convection. However, the excursion region is fully bypassed during the cooling mode, which is why differences between heating and cooling modes are described as ‘incipience hysteresis.’ One reason for lack of excursion during the cooling mode is abundance of vapor from the fully developed nucleate boiling region, which serves to provide larger vapor embryos in active cavities and activate neighboring initially inactive cavities [35]. These phenomena have been reported by many investigators [29], [36], [37], [38], [39], who cautioned against their damaging effects for temperature sensitive devices. Potential damage to these devices is encountered in two different ways, first due to surface overheating just prior to incipience, and immediately following incipience due to the large temperature drop; the latter may induce ‘thermal shock’ to the device. It should be noted that these incipience effects are observed with both bare and enhanced surfaces. Following is a more detailed look into the surface temperature drop immediately following incipience.

Clearly, the large temperature drop that accompanies boiling incipience of low contact angle coolants is the product of the large, rapid change in heat transfer coefficient as the heat transfer mode transitions from natural convection to nucleate boiling. For a small, fairly isothermal surface undergoing both natural convection and nucleate boiling over separate areas A_nc and A_b, respectively, total heat transfer is the net result of both heat transfer modes,

$$\frac{q^{″}}{\mathrm{\Delta }{T}_{\mathit{sat}}}=hA=h_{\mathit{nc}}{A}_{\mathit{nc}}+h_b{A}_b,$$

where h is the overall heat transfer coefficient, and h_nc and h_b are the heat transfer coefficients associated with natural convection and nucleate boiling, respectively. By defining the boiling area fraction as A⁺ = A_b/A, the heat transfer coefficient can be expressed as

$$h=h_{\mathit{nc}}\left(1-A^{+}\right)+h_b{A}^{+}.$$

Notice in Eq. (3) how as A⁺ grows gradually toward unity (fully developed nucleate boiling region), the overall heat transfer coefficient h will likewise increase to the value of h_b. The transient increase in A⁺ upon incipience is influenced by surface size and pattern of boiling activation. Small heaters incur large jumps in A⁺ since the area influenced by boiling from a single site can occupy a significant fraction of the surface area [8]. Rapid patterns of activation (e.g., due to surface orientation or surface enhancement) can cause large swings in A⁺ and therefore result in considerable temperature drop.

Another contributor to the magnitude of incipience temperature drop is conduction within the heating surface. Foil heaters, such as those used by Park and Bergles [40], allow little conduction across the heat transfer surface. Consequently, temperature drop is highly localized, and the large change in local heat transfer coefficient is sensed as a large temperature fallback. In contrast, thick heaters allow significant conduction within the heating surface, which dampens localized effects of activation of a few surface cavities. Rapid activation of the entire surface may still induce a measurable temperature drop, but thermal shock is damped by increased thermal capacitance for thick heaters.

Another obvious limitation for pool boiling in dielectric fluids is relatively low critical heat flux (CHF). This is readily apparent when comparing CHF from a bare surface in a fluorochemical to that in water at atmospheric pressure. According to the Zuber model [41], [42], [43], CHF for saturated pool boiling in water is 110.4 W/cm², whereas that for FC-72 is only 15.24 W/cm². Inferior CHF for FC-72 is a direct result of the relatively poor thermophysical properties of FC-72, especially latent heat of vaporization and surface tension. Given the safety requirement of maintaining heat flux below 70% of CHF [44], the upper operational limit for device heat flux is reduced even further. Low CHF is especially problematic for compact electronic cooling systems employing a small boiler. Here, confining walls of the small boiler may interrupt liquid replenishment of the heating surface during vigorous boiling, thereby compromising CHF even further.

Methodologies that have been proposed to improve nucleate boiling heat transfer can be classified into two major groups: active and passive [45]. The active techniques require external power, and include mechanical mixing, surface and/or liquid rotation, vibration, suction or injection, and addition of external electrostatic or magnetic fields, all of which have been proven effective at intensifying heat transfer [46]. However, active enhancement techniques are costly and difficult to implement in compact cooling enclosures, such as those used for electronics cooling. In contrast, passive techniques require no direct application of external power, and include modifying coolant properties and/or surface roughness and shape, or using finned surface attachments to increase surface area.

Compared to flow boiling, pool boiling offers few heat transfer enhancement options. Increasing flow velocity in flow boiling improves both single-phase and two-phase heat transfer and postpones CHF by sweeping bubbles away from the surface before they coalesce into an insulating vapor blanket [47]. In pool boiling, however, relatively weak fluid motion is induced by natural convection prior to the onset of boiling and, thereafter, by buoyancy and bubble agitation. Unfortunately, the induced flow cannot be easily controlled. Therefore, attempts at enhancing pool boiling performance are limited mostly to two major methodologies: (a) modifying the fluid itself and/or operating conditions (e.g., saturation pressure and subcooling), and (b) modifying the heating surface. Surface modification can be achieved on a variety of scales: ‘macroscale,’ ‘microscale,’ and ‘nanoscale,’ which are aimed at increasing heat transfer area, increasing nucleation site density, and improving capillary wicking effects, respectively. There are also ‘hybrid-scale’ methods that combine the cooling merits of different surface modification scales in pursuit of superior cooling performance.

Temperature-sensitive devices such as electronic and power devices impose stringent operational limits to implementing a thermal management system employing pool boiling. Most important among these limits are (1) maximum operating temperature (e.g., 125 °C for current high performance electronic and power devices, imposed by both materials and reliability concerns), and (2) CHF. Fig. 5 illustrates how the boiling system must be configured to meet both limits. Shown first are boiling characteristics for a bare (unenhanced) surface, where neither limit is met. Also shown are characteristics for an enhanced albeit unacceptable surface, where, despite superior CHF, maximum surface temperature is exceeded at the maximum operating heat flux. A third case is also shown, where both temperature and CHF limits are properly met. To achieve the desired cooling performance indicated by the third case, it is desired to reduce wall superheat and increase CHF simultaneously.

One application where pool boiling might benefit from surface enhancement is thermal management in space systems. Absence of gravity is known to greatly compromise cooling effectiveness by triggering CHF at unusually low heat fluxes [48], [49], [50], [51]. Without significant additional enhancement, pool boiling is unlikely to become a viable cooling option for these applications. Here, use of structured surfaces might aid in both increasing surface area and breakup of large coalescent vapor bubbles, thereby ameliorating CHF.

Aside from meeting the two above operational limits, boiling enhancement is employed to achieve other goals as well. Overall, pool boiling enhancement is used to achieve one or more of the following goals: (a) initiating nucleate boiling at lower surface temperature, (b) reducing and/or eliminating incipience temperature excursion, especially for low contact angle coolants, (c) reducing surface temperature (i.e., increasing the heat transfer coefficient associated with nucleate boiling), and (d) increasing CHF to accommodate higher surface heat fluxes.

Several review articles have been published in the past, which address boiling heat transfer enhancement by means of surface modification. These articles either address enhancement of pool boiling along with other heat transfer modes, or are dedicated entirely to pool boiling. Examples of the first category include works by Bergles and co-workers documenting 508 papers and reports [45] and 59 U.S. patents [52] relating to both pool boiling and flow boiling prior to 1980. Bhavnani et al. [53] and Shojaeian and Koşar [54] also reviewed enhancement literature related to both pool and flow boiling using micro/nanostructured surfaces. McCarthy et al. [55], Kim et al. [56], and Attinger et al. [57] published review articles addressing practical concerns with boiling enhancement, including both materials and methods used to fabricate micro/nanostructured surfaces. Examples of reviews dedicated entirely to pool boiling enhancement include an article by Honda and Wei [58] addressing surface microstructures developed prior to 2004 to improve cooling performance for electronic devices. Lu and Kandlikar [59] reviewed pool boiling enhancement using nanoscale surface modification techniques, and follow-up work by Patil and Kandlikar [60] reviewed techniques for manufacturing porous surfaces for pool boiling applications. A recent article by Mori and Utaka [61] reviewed techniques that have been used to enhance pool boiling CHF, but did not address enhancement of heat transfer coefficient within the nucleate boiling region.

The present paper is a follow-up to a series of reviews by the present authors addressing several fundamental two-phase mechanisms, including fluid mechanics of liquid drop impact on a liquid film [62] and on a heated wall [63], spray cooling in single-phase regime, nucleate boiling regime, and CHF [4], and in high temperature boiling regimes and quenching applications [5], mechanisms and models of pool boiling CHF [64], and pool boiling enhancement techniques using additives and nanofluids [65].

The present review is intended to provide a very comprehensive assessment of passive pool boiling heat transfer enhancement using surface modification. Included are potential problems associated with cooling temperature sensitive electronic and power devices and key challenges for pool boiling enhancement. Key topics discussed are boiling enhancement at different surface modification scales, including macroscale, microscale, and nanoscale, with feature sizes larger than 1 mm, in the range of 1–1000 μm, and smaller than 1 μm, respectively. Also included are hybrid-scale enhancement methods, which combine two or more of the enhancement scales just mentioned. Each of the topics presented also includes description of fabricating techniques used in the surface modification. This paper also discusses novel concepts for separating liquid and vapor paths, and pumpless cooling loops. The review is concluded with recommendations concerning future work that is needed to address poorly understood mechanisms or contradictory findings. It should be noted that this review is limited to surface enhancement applied to planner heat dissipating surfaces, and excludes tubular surfaces.